Question-Driven Design Process for Explainable AI User Experiences

User needs should drive XAI design choices: Rather than starting with available technical tools, the design process should begin by identifying what specific questions users actually need answered about an AI system. The paper demonstrates this through a healthcare case study where understanding clinicians' information needs shaped which explainability techniques were ultimately selected.

There's a gap between XAI technical capabilities and practical design: The paper identifies that while explainable AI researchers have developed many techniques, designers struggle to translate these into user experiences that actually help people understand AI decisions. This process bridges that gap by creating a mapping guide connecting user questions to appropriate XAI techniques.

Explainability is not one-size-fits-all: Different users have different questions about why an AI made a particular decision. The question-driven approach recognizes that a healthcare worker might need different explanations than a patient or a regulator, requiring tailored XAI solutions rather than generic explanations.

Better collaboration between designers and engineers is needed: By using user questions as "boundary objects," the framework helps designers and AI engineers communicate more effectively about which explainability techniques will actually serve end users, rather than defaulting to the most technically sophisticated options.

Evaluation should match the original user questions: The paper emphasizes that XAI systems should be tested based on whether they successfully answer the specific user questions they were designed to address, ensuring accountability throughout the design process.